[squid-users] cache digest utilization

From: Kevin Quinlan <kevin.quinlan@dont-contact.us>
Date: Thu, 25 Oct 2001 22:57:55 +0100

Apologies for the earlier incomplete message, a quirk in MS Outlook I am
afraid.

Here is the completed message:

We have a "farm" of 3 squid proxy servers running 2.4 STABLE2.

They are configured to use digests to try and share cache data, for the
purposes of describing this issue I will talk about 2 of the machines as
they are configured identically, the 3rd server has different RAM and disk
capacity.

The machines run FreeBSD 4.1, they have 1Gb RAM and 2 18Gb disks forming the
cache runnning at about 83% capacity.

The disks are set up thus:

cache_dir diskd /cache0/squid 14000 16 256
cache_dir diskd /cache1/squid 14000 16 256

Peers are configured thus:

cache_peer proxy3.ctxuk.citrix.com sibling 3128 3130

digest parameters are currently all defaults:

# digest_bits_per_entry 5
# digest_rebuild_period 1 hour
# digest_rewrite_period 1 hour
# digest_swapout_chunk_size 4096 bytes
# digest_rebuild_chunk_percentage 10

Looking at the peer selection algorithms for this server, I see

peer digest structure:
proxy3.ctxuk.citrix.com digest: size: 1685126 bytes
         entries: count: 47199 capacity: 2696201 util: 2%
         deletion attempts: 0
         bits: per entry: 5 on: 187428 capacity: 13481008 util: 1%
         bit-seq: count: 369446 avg.len: 36.49

and both servers have run at this consistently low cache utilization for the
last month.

(It is actually much the same for the 3rd server which has about half the
RAM and half the disk)

I have read the FAQ and searched back though this list and read some of the
book, but I have not found an explanation of how to tune the digest
generation if the algorithm produces bad results as my servers seem to be
doing.

I did try to change the bits per entry setting, but after a while the server
set it back to 5 anyway.

I would appreciate some pointers. I must have configured something
sub-optimally but I cannot see what.

I am not sure that thw whole config is relevant, but here is some more
information on the cache size:

cache_mem 48 MB
cache_swap_low 90
cache_swap_high 95
maximum_object_size 65535 KB
minimum_object_size 0 KB
maximum_object_size_in_memory 24 KB
ipcache_size 1024
ipcache_low 90
ipcache_high 95
fqdncache_size 1024
cache_replacement_policy heap LFUDA
memory_replacement_policy heap LFUDA

Thanks

Kevin Quinlan
Received on Thu Oct 25 2001 - 15:57:59 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:03:09 MST